JP6522149B2 - System and method for predicting perfusion injury from physiological characteristics, anatomical characteristics, and patient characteristics - Google Patents

Description

RELATED APPLICATIONS This application claims priority under 35 USC 119119 (e) to US Provisional Patent Application No. 62 / 142,158 filed on April 2, 2015, the contents of which are incorporated herein by reference. , Which is incorporated herein by reference in its entirety.

  Various embodiments of the present disclosure generally relate to disease assessment, treatment regimens, and related methods. More specifically, certain embodiments of the present disclosure relate to systems and methods for estimating perfusion of target tissue.

  Coronary artery disease is a common disease that affects millions of people. Coronary artery disease may cause lesions, such as stenosis (abnormal narrowing of blood vessels), in blood vessels that provide blood to the heart. As a result, blood flow to the heart may be limited. Patients suffering from coronary artery disease may experience chest pain, called chronic stable angina during physical exercise or unstable angina when the patient is at rest. More serious manifestations of the disease may lead to myocardial infarction or heart attack. For coronary artery disease, including both medical treatment (eg, statins) or surgical alternatives (eg, percutaneous coronary intervention (PCI) and coronary artery bypass graft surgery (CABG)) Significant progress has been made in treatment. Invasive assessment is commonly used to assess the type of treatment that a patient may receive. However, indirect or non-invasive evaluations for devising patient treatment have been investigated and developed.

  Heart disease is usually considered to be due to vascular disease, in particular, narrowing of the blood vessel to affect blood flow or occlusion inside the blood vessel lumen. One way to measure this degree of narrowing or occlusion is by perfusion scanning. The reason is that perfusion is the flow of blood through the vascular network (eg, arteries, capillaries etc.). Currently, perfusion scans can be expensive and can expose the patient to unwanted radiation. As such, there is a desire to use available patient information to estimate perfusion in a given target tissue, and the estimated perfusion data is used to tell the physician how to interpret the image, It may simulate well-known scan types, such as single photon emission computed tomography (SPECT) or positron emission tomography (PET). Furthermore, there is a desire to improve the treatment of cardiovascular disease by better assessing the severity of the cardiovascular disease.

  The foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.

  According to one aspect of the present disclosure, systems and methods are disclosed for using the available information to estimate perfusion of target tissue to guide diagnosis or treatment of cardiovascular disease.

  One method is to receive a patient-specific vascular model and a patient-specific tissue model of the patient's anatomy, one or more of the patient from the blood vessel or tissue model in one or more physiological states of the patient. Extracting patient-specific physiological parameters of the patient, estimating characteristics of blood perfusion through the patient-specific tissue model using patient-specific physiological parameters, and displaying estimated characteristics of blood perfusion Including output to

  According to another embodiment, a system for estimating patient-specific perfusion comprising: a data storage device storing instructions for determining patient-specific characteristics of blood perfusion; and a processor Receiving the patient-specific vascular model and the patient-specific tissue model of the patient's anatomical structure, and one of the patient from the blood vessel or tissue model in one or more physiological states of the patient. Extracting one or more patient-specific physiological parameters; estimating characteristics of blood perfusion through the patient-specific tissue model using the patient-specific physiological parameters; Outputting the characteristic to a display, the system being configured to execute instructions implementing the method.

According to another embodiment, a non-transitory computer readable medium for use on a computer system including computer executable programming instructions for estimating patient specific characteristics of blood perfusion, the method comprising: Receiving a patient-specific vascular model and a patient-specific tissue model of the anatomical structure of the patient, one or more patient-specific physiological models of the patient from the vascular or tissue model in one or more physiological states of the patient Extracting parameters, estimating patient-specific characteristics of blood perfusion through a patient-specific tissue model using patient-specific physiological parameters, and outputting the estimated characteristics of blood perfusion to a display , Non-transitory computer readable medium.
The present specification also provides, for example, the following items.
(Item 1)
A computer-implemented method for estimating patient-specific characteristics of blood perfusion through a patient's anatomy, comprising:
Receiving a patient-specific vascular model of the patient's anatomical structure and a patient-specific tissue model,
Extracting one or more patient-specific physiological parameters of the patient from the vessel or tissue model in one or more physiological states of the patient;
Go through the patient-specific tissue model using a machine learning algorithm trained to map patient perfusion characteristics to the patient-specific physiological parameters and patient-specific anatomical geometry and physiological parameters Estimating the characteristics of said perfusion of blood;
The computer-implemented method comprising outputting the estimated property of the perfusion of blood to an electronic display.
(Item 2)
The computer-implemented method of claim 1, wherein the patient-specific physiological parameter comprises an estimated or measured blood flow to multiple regions of the blood vessel or tissue in one or more physiological conditions.
(Item 3)
The patient specific physiological parameters include, but are not limited to, vessel size, vessel shape, degree of vessel flexion, vessel length, vessel thickness, estimated territory of the perfusion of tissue or blood within the vessel network, or a combination thereof , A computer implemented method according to item 1.
(Item 4)
The physiological condition may include one or more of resting patient condition, congestion condition, exercise condition, postprandial condition, gravity condition, emotional condition, hypertension condition, drug taking condition, or a combination thereof. A computer-implemented method as described in.
(Item 5)
The computer-implemented method of claim 1, wherein the perfusion characteristics of blood include one or more of coronary flow reserve ratio, flow magnitude, flow direction, or a combination thereof.
(Item 6)
The patient-specific physiological parameter may be one or more of a target tissue or vessel model in one or more physiological conditions, including one or more of local average intensity, texture characteristics, standard images, or a combination thereof. The computer-implemented method of claim 1, comprising an image characteristic.
(Item 7)
The computer-implemented item of claim 1, wherein the patient specific physiological parameter further comprises a plurality of secondary characteristics including one or more of patient characteristics, target tissue disease characteristics, electromechanical measurements, or a combination thereof. Method.
(Item 8)
The computer-implemented item of claim 1, wherein estimating characteristics of the perfusion of blood through the patient-specific tissue model comprises comparing blood flow within the target tissue in one or more physiological states. Method.
(Item 9)
The patient-specific vascular model and patient-specific tissue model of the patient's anatomical structure include coronary and myocardial, cerebrovascular and brain, peripheral vascular and muscle, hepatic vascular and liver, renal vascular and kidney models. 10. A computer-implemented method according to claim 1, including one or more of: a visceral vascular model and a gut or a vascular model having any target organ and blood vessels supplying the target organ.
(Item 10)
Adjusting the patient-specific physiological parameter based on the estimated characteristic of the perfusion of blood;
The computer-implemented method of claim 1, comprising simulating blood flow characteristics using the estimated characteristics of the perfusion of blood and adjusted physiological parameters.
(Item 11)
Receiving one or more desired perfusion characteristics at one or more locations within the target tissue;
Comparing the estimated property of the perfusion of blood at the one or more locations in the target tissue with the desired perfusion characteristic at the one or more locations;
The computer-implemented method of claim 1, further comprising modifying one or more of the physiological parameter, a patient-specific anatomical model of the blood vessel or target tissue based on the comparison.
(Item 12)
One or more virtual perfusion images designed to simulate a SPECT or PET scan in one or more physiological conditions using the one or more estimated characteristics of the perfusion of blood Estimating, and
The computer-implemented method of claim 1, further comprising outputting the one or more virtual perfusion images to an electronic storage device.
(Item 13)
A system for estimating patient-specific perfusion, comprising
A data storage device storing instructions for determining patient specific characteristics of blood perfusion;
A processor, the processor being a method,
Receiving a patient-specific vascular model of the patient's anatomical structure and a patient-specific tissue model;
Extracting one or more patient-specific physiological parameters of the patient from the blood vessel or tissue model in one or more physiological states of the patient;
Estimating the characteristics of the perfusion of blood through the patient-specific tissue model using the patient-specific physiological parameters;
Outputting the estimated property of the perfusion of blood to an electronic display, the system configured to execute the instructions to perform the method.
(Item 14)
The patient specific physiological parameters may, in one or more physiological conditions:
An estimated or measured blood flow to a plurality of regions of the blood vessel or tissue;
Anatomical characteristics,
Image characteristics
Patient characteristics,
Target tissue disease characteristics,
Electromechanical measurements, or
The combination,
14. A system according to item 13, comprising one or more of
(Item 15)
The physiological condition may be one or more of resting patient condition, congestion condition, exercise condition, postprandial condition, gravity condition, emotional condition, hypertension condition, drug taking condition, or a combination thereof. System.
(Item 16)
14. A system according to item 13, wherein the perfusion characteristic of blood comprises one or more of coronary flow reserve ratio, flow size, flow direction, or a combination thereof.
(Item 17)
14. The system according to claim 13, wherein estimating the property of the perfusion of blood through the patient specific tissue model comprises comparing blood flow within the target tissue in one or more physiological states.
(Item 18)
The patient-specific vascular model and patient-specific tissue model of the patient's anatomy are:
Coronary blood vessel model and myocardium,
Cerebrovascular model and brain,
Peripheral vascular model and muscle,
Liver blood vessel model and liver,
Renal vascular model and kidney,
Visceral vascular model and intestine,
Or a blood vessel model having an arbitrary target organ and a blood vessel supplying the target organ.
14. A system according to item 13, comprising one or more of
(Item 19)
Adjusting the patient-specific physiological parameter based on the estimated characteristic of the perfusion of blood;
14. A system according to item 13, further comprising simulating blood flow characteristics using the estimated characteristics of the perfusion of blood and adjusted physiological parameters.
(Item 20)
A non-transitory computer readable medium for use on a computer system comprising computer executable programming instructions for estimating patient specific characteristics of blood perfusion, said method comprising
Receiving a patient-specific vascular model of the patient's anatomical structure and a patient-specific tissue model,
Extracting one or more patient-specific physiological parameters of the patient from the vessel or tissue model in one or more physiological states of the patient;
Estimating the characteristics of the perfusion of blood through the patient-specific tissue model using the patient-specific physiological parameters;
Outputting the estimated property of the perfusion of blood to an electronic display;
Said non-transitory computer readable medium.

  Additional objects and advantages of the disclosed embodiments will be set forth in part in the description that follows, and will in part be apparent from the description or practice of the disclosed embodiments. May be learned by The objects and advantages of the disclosed embodiments will be realized and attained by the elements and combinations particularly pointed out in the appended claims.

  It is understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the claimed detailed embodiments.

  BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate various exemplary embodiments and, together with the description, serve to explain the principles of the disclosed embodiments.

FIG. 1 is a block diagram of an exemplary system and network for predicting perfusion to guide diagnosis or treatment of cardiovascular disease, according to an exemplary embodiment of the present disclosure. FIG. 1 is a block diagram of a general method of estimating perfusion, according to a general embodiment of the present disclosure. FIG. 7 is a block diagram of an exemplary method of estimating perfusion, according to an exemplary embodiment of the present disclosure. FIG. 8 is a block diagram of an exemplary method of estimating perfusion using machine learning, in accordance with an exemplary embodiment of the present disclosure, for determining an estimated amount of tissue perfusion, respectively, at step 208 in FIGS. 2 and 3; Or similarly disclosed methods of practicing 322. FIG. 5 is a block diagram of the method disclosed in FIG. 4 and further, using machine learning, in accordance with an exemplary embodiment of the present disclosure, blood vessel geometry and patient specific physiological parameters (eg, anatomical information, 2 An exemplary method for estimating patient-specific perfusion characteristics from the following information, and blood supply information) is likewise disclosed.

  Reference will now be made in detail to exemplary embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings to refer to the same or like parts.

  Coronary artery disease is a common disease in which blood flow to the heart may be limited. Although significant progress has been made in the treatment of coronary artery disease, the treatment is often out of place or in excess. For example, patients often undergo invasive surgical procedures or perfusion scans, which may be expensive and / or may expose the patient to unnecessary radiation. Patients sometimes receive treatments that do not have the potential to change their condition. In certain situations, the patient is further subjected to a treatment that ultimately worsens his condition. As such, there is a need to accurately assess the severity of cardiovascular disease and / or predict perfusion to aid in selecting a course of treatment.

  Cardiovascular disease is linked to vascular disease and may mean vasoconstriction or occlusion. A cardiac perfusion scan can measure the amount of blood in the myocardium at different physiological conditions. "Physiological state" may refer to resting patient state, congestion state, exercise state, postprandial state, gravity state, emotional state, hypertension state, drug taking state, or a combination thereof. A perfusion scan is often performed to determine what may be causing chest pain, and also to determine if the tissue of the heart is being provided with the proper flow of blood, or the myocardium But determine how much damage from a heart attack.

  During the scan, an image of the heart is generated after the radioactive tracer has been intravenously administered to the patient. The radioactive tracer travels through the blood and into the myocardium. As the tracer travels through the myocardium, tissue with sufficient blood flow absorbs the tracer. Tissues that do not absorb the tracer may not receive enough blood or may be damaged by a heart attack. Two sets of images may be generated during a cardiac perfusion scan. The resting image may then be compared to a stress or non-resting image to determine the level of perfusion in the target tissue. "Target tissue" may refer to a tissue and / or organ from which blood supply and / or perfusion characteristics may be estimated.

  Thus, an understanding of perfusion in the target tissue may be clinically important. An understanding of perfusion may improve the assessment of disease severity or appropriateness of treatment. The present disclosure more accurately assesses the severity of vascular disease in different physiological situations, where perfusion may be difficult to measure, and / or by estimating perfusion in situations where it may be difficult to measure The use of perfusion measurements can benefit the patient and the physician. The generated perfusion images can simulate images generated from well-known scan types such as PET and / or SPECT. The simulated image can provide ease of interpretation and understanding to a physician trained to interpret PET and / or SPECT images. For the purposes of the present disclosure, "patient" may be any individual or person for whom diagnosis or treatment analysis is to be performed, or any individual associated with diagnosis or treatment analysis of one or more individuals. Or it may refer to a person.

  While FIG. 1 provides an abstract view of the systems and networks of the present disclosure, FIG. 2 shows a general embodiment of a method for estimating one or more characteristics of blood perfusion, and FIG. , Place more specific embodiments. Further, FIGS. 2 and 3 together disclose the step of determining an estimated amount of blood perfusion through the tissue. FIG. 4 illustrates an exemplary method of performing the step of determining an estimate of one or more characteristics of perfusion of blood through tissue using a trained machine learning algorithm. FIG. 5 considers the method disclosed in FIG. 4 in more detail.

  With reference now to the figures in more detail, FIG. 1 shows a block diagram of an exemplary system 100 and network for estimating cardiac perfusion to guide diagnosis or treatment of cardiovascular disease according to an exemplary embodiment. . In particular, FIG. 1 shows a plurality of physicians 102 and third party providers 104, any of which may be electronic, such as the Internet, through one or more computers, servers, and / or handheld mobile devices. It may be connected to the network 101. The physician 102 and / or the third party provider 104 may create or otherwise obtain an image of the anatomical structure of one or more patients. The physician 102 and / or third party provider 104 may similarly obtain any combination of patient specific information such as age, medical history, blood pressure, blood viscosity, patient activity, or exercise level etc. The physician 102 and / or the third party provider 104 may transmit anatomical images and / or patient specific information to the server system 106 via the electronic network 101. Server system 106 may include a storage device for storing images and data received from physician 102 and / or third party provider 104. The server system 106 may also include processing devices for processing the images and data stored on the storage device.

  FIG. 2 shows a general embodiment of an exemplary method 200 for estimating cardiac perfusion to guide diagnosis or treatment of cardiovascular disease. The method of FIG. 2 may be implemented by server system 106 based on information, images, and data received from physician 102 and / or third party provider 104 through electronic network 101.

  In one embodiment, step 202 may include receiving a patient-specific vascular model of a patient's anatomical structure and a patient-specific target tissue model stored in an electronic storage medium of storage system 106. An "electronic storage medium" may include a hard drive, a network drive, a cloud drive, a mobile phone, a tablet, or the like, which may or may not be affixed to the display screen. But it is not limited to it. In particular, receiving a patient-specific vascular model and a patient-specific target tissue model may include generating a patient-specific blood vessel model and / or a patient-specific target tissue model at server system 106 or an electronic network (eg, Receiving the patient-specific vascular model and / or the patient-specific target tissue model through the electronic network 101). The patient-specific vascular model and the patient-specific target tissue model may include a cardiovascular model of a particular patient. In one embodiment, the blood vessel model and the target tissue model are one or more available imaging or scanning modalities (eg, computed tomography (CT) scans and / or magnetic resonance imaging (MRI) It may be derived from the image of the person who is acquired by). For example, step 202 may include receiving CT and MRI images of a human heart. Step 202 may further include generating a patient-specific cardiovascular model for the particular person from the received images. Electronic storage media may include, but is not limited to, a hard drive, a network drive, a cloud drive, a mobile phone, or the like.

  In one embodiment, steps 204 and 206 may include receiving or calculating one or more patient specific physiological parameters. These patient specific physiological parameters may be received or calculated from the received vessel model and / or target tissue model. These patient-specific physiological parameters include anatomical characteristics, imaging characteristics, and secondary characteristics associated with the patient and / or patient anatomy (eg, patient characteristics, disease load characteristics, and electromechanical characteristics). Measurement values may be included. Patient-specific physiological parameters also include parameters related to blood circulation, including estimated amounts of blood supply and / or blood flow characteristics for each area of target tissue, under one or more physiological conditions. May be.

  In particular, step 204 receives or calculates one or more anatomical characteristics, image characteristics, patient characteristics, disease load characteristics, and / or electromechanical measurements under one or more physiological conditions. May be included. One instance of a physiological condition may be a resting condition. Another physiological condition may be a physiological condition other than the resting condition or an "active" physiological condition. Active physiological conditions may include congestion, various exercise levels, postprandial, position (eg, supine- upright), gravity (eg, G-force, zero gravity, etc.), or a combination thereof. In one embodiment, patient-specific physiological parameters may be obtained from sources other than vascular models and / or target tissue models.

  In one embodiment, step 206 receives or calculates estimated blood supply and / or estimated blood flow characteristics for each blood vessel in each area or network of target tissue under one or more physiological conditions. May be included. These estimates may be based on measurements (e.g. by measuring through imaging) or by estimation of the supplied blood in a resting state (e.g. based on 3D simulations, 1D simulations or learned relationships) It is also good.

  In one embodiment, step 208 may include using linked prior information to determine or output an estimated amount of perfusion in the blood vessel of the blood vessel model or an area in the target tissue model. The connected prior information may be for one or more received patient-specific physiological parameters (eg, received or calculated medical image characteristics, anatomical characteristics, target tissue) determined at steps 204 and 206. Blood supply, blood flow characteristics, patient characteristics, disease load characteristics, electromechanical measurements, etc.) may also be referred to. In one embodiment, determining an estimated amount of perfusion comprises determining an estimated amount of blood supply in one or more blood vessel locations of a human blood vessel model while the person is at a given physiological condition. May be included. This determination may also be based on measurement of blood flow (e.g. by imaging) or estimation of blood flow in resting state (e.g. based on 3D simulation, 1D simulation or learned relationships) It is also good. In one embodiment, step 208 may include calculating an estimate of perfusion territory of the target tissue associated with the vascular model. This estimate may be determined by using a nearest neighbor (e.g., Voronoi diagram) approach that assigns a location in the target tissue to the closest supply vessel in the vessel model. Estimators may likewise be determined using microvessel estimation techniques from anatomical models, for example by using a constrained constructive optimization approach. In one embodiment, step 208 may be performed by a processor. The processor may estimate perfusion at one or more locations of the target tissue in the blood vessel model at one or more physiological conditions by machine learning. In one embodiment, step 208 may further include outputting an estimated amount of perfusion to an electronic storage medium (eg, hard disk, network drive, portable disk, smartphone, tablet, etc.) and / or a display screen. In one embodiment, the output perfusion estimate may be displayed in 2D or 3D in grayscale or in color. The calculated perfusion estimate may be superimposed on the anatomical model of the target tissue and / or may be superimposed on the image of the target tissue.

  In one embodiment, step 210 may include estimating virtual perfusion designed to simulate a SPECT or PET scan in one or more physiological conditions. In one embodiment, the estimation may be performed by modeling the concentration of contrast agent given by the perfusion estimate. In another embodiment, the estimation may include performing a Monte Carlo simulation that estimates collimation of photons or protons at virtual collimator locations. Using collimator estimation, SPECT or PET images may be reconstructed using standard tomographic techniques. The estimated virtual perfusion image may be stored on an electrical storage medium and / or displayed on a monitor.

  FIG. 3 shows an exemplary embodiment of a method 300 for estimating cardiac perfusion to guide the diagnosis or treatment of cardiovascular disease. The method of FIG. 3 may be implemented by server system 106 based on information, images, and data received from physician 102 and / or third party provider 104 through electronic network 101.

  In one embodiment, step 302 may include a patient-specific blood vessel model of at least target tissue of a patient anatomical structure in the electronic storage medium of storage system 106. In particular, receiving a patient-specific anatomical model includes generating the anatomical model at server system 106 or receiving the anatomical model via an electronic network (eg, electronic network 101) May be. The patient-specific anatomical model may include a particular human cardiovascular model. In one embodiment, the anatomical model may be derived from an image of the person acquired by one or more available imaging or scanning modalities (eg, CT scan and / or magnetic resonance imaging). For example, step 302 may include receiving a CT and / or MRI image of a human heart. Step 302 may further include generating a patient-specific cardiovascular model for a particular person from the received images. Electronic storage media may include, but is not limited to, a hard drive, a network drive, a cloud drive, a mobile phone, a tablet, or the like.

  In one embodiment, step 304 may include calculating one or more estimators of the anatomical characteristics of the target tissue. Anatomical properties include, but are not limited to, vessel size, vessel shape, tortuosity, thickness, and / or estimated perfusion territory within the vascular tissue or vascular network. This calculation may be based on measurements (e.g. by measuring through imaging) or estimation of anatomical properties in resting state (e.g. based on 3D simulations, 1D simulations or learned relationships) It is also good.

  In one embodiment, step 306 may include receiving one or more image characteristics of the target tissue. Image characteristics may be obtained from CT scan images, MRI images, ultrasound images, PET images, or SPECT images. The image may capture a vascular model at one or more physiological conditions (eg, rest, stress, active). Image characteristics of the target tissue or blood vessel may be received or calculated at one or more locations in the blood vessel model. Image characteristics include local average intensity at one or more image resolutions, difference in average intensity (e.g., calculated by wavelet based using Haar wavelets), texture characteristics (e.g., Halalic texture features), and Any standard image feature may include, but is not limited to, any standard image feature, including histograms, gradients, SIFTs, or steerable filters and the like.

  In one embodiment, step 308 may include receiving a patient characteristic. Patient characteristics may include, but are limited to: age, gender, smoking history, height, weight, diabetes status, hypertension status, ethnic status, family history, blood type, history of drug use, and / or genetic history I will not. Patient characteristics may be obtained via the electronic network 100 or from the patient's doctor 102 or from a third party provider 103.

  In one embodiment, step 310 may include receiving a blood vessel or target tissue disease property. Target tissue disease characteristics include the presence and degree of plaque accumulation in arteries, the presence of plaque characteristics (eg, microcalcifications, low absorption plaques, napkin ring signatures, positive remodeling), patient level or vascular level calcium score, tissue survival It may include, but is not limited to, information, vessel wall motion, vessel wall thickness, and / or ejection fraction information.

  In one embodiment, step 312 may include receiving an electromechanical measurement. The electro-mechanical measurements may include, but are not limited to electrocardiography (ECG) measurements or invasive electrophysiology (EP) measurements.

  In one embodiment, step 314 may include calculating an estimate of the blood supply for each area of target tissue under one or more physiological conditions. One instance of the first physiological condition may be a resting condition. This calculation may be based on measurements (e.g. by measuring through imaging) or at rest (e.g. based on three-dimensional (3D) simulation, one-dimensional (1D) simulation or learned relationships) It is also possible to estimate the blood supply in the condition. Another physiological condition may be a physiological condition other than the resting condition or an "active" physiological condition. One instance of such a physiological condition may involve congestion. Other non-resting physiological states may include various levels of movement, postprandial, position (eg, supine- upright), gravity (eg, G-force, zero gravity, etc.).

  In one embodiment, step 316 may include calculating one or more blood flow characteristics of the target tissue. In one embodiment, the blood flow characteristics may include, but are not limited to, flow reserve ratio (FFR), flow direction, and / or flow size, but not limited to, blood flow to target tissue It is determined by the estimate of In one embodiment, blood flow characteristics are calculated using invasive measurements (eg, invasive FFR, thrombosis in myocardial infarction (TIMI), or microspheres), a blood flow simulation model (eg, Including 3D or 1D fluid simulation models, calculations, or calculations using TAFE), calculations using imaging properties (eg TAG or CCO), or calculations using machine learning estimators of blood supply based on anatomical or imaging features It may be calculated by several means, not limited thereto. In one embodiment, step 316 may include calculating an estimate of blood flow within the perfusion territory of the target tissue associated with the vascular model. This estimator may be determined by using the nearest neighbor (eg, Voronoi diagram) approach of assigning a location in the target tissue to the closest supply vessel in the vessel model. Estimators may also be determined using microvessel estimation techniques from anatomical models, for example by using a constrained constructive optimization approach. In one embodiment, step 316 may be performed by a processor. The processor may estimate perfusion at one or more locations of the target tissue in the blood vessel model at one or more physiological conditions by machine learning.

  In one embodiment, step 318 may include using a processor to calculate an estimated amount of perfusion at one or more locations in the target tissue associated with the vascular model. The estimated amount of perfusion may be patient-specific physiological parameters (eg, image characteristics, anatomical characteristics, estimated blood supply, estimated blood flow characteristics, estimated perfusion territory, patient characteristics, disease load characteristics, and / or electricity. One or more of the mechanical measurements) may be used to calculate one or more physiological conditions. In one embodiment, this calculation includes known perfusion characteristics and known patients including, but not limited to, imaging characteristics, anatomical characteristics, estimated perfusion territory, disease loading characteristics, and / or electromechanical measurements. It may be implemented by training a machine learning algorithm using a database of patients having unique physiological parameters. In one embodiment, step 318 may be implemented using a processor.

  In one embodiment, step 320 may include outputting an estimated amount of perfusion to an electronic storage medium (eg, hard disk, network drive, portable disk, smart phone, tablet, etc.) and / or a display screen. In one embodiment, the output perfusion estimate may be displayed in 2D or 3D in grayscale or in color. In one embodiment, the output perfusion estimate may be superimposed or superimposed on an anatomical model of the target tissue, and / or may be superimposed on an image of the target tissue It may be good or superimposed. In one embodiment, this determination may be performed by training a machine learning algorithm using a database of patients with known perfusion characteristics and known patient-specific physiological parameters. In one embodiment, the perfusion estimate may be used to estimate a virtual perfusion image designed to simulate a SPECT or PET image in one or more of the patient's physiological conditions. In one embodiment, the virtual perfusion image may be stored on an electronic storage device and / or output to a display. The estimation of the virtual perfusion image may be performed by modeling a contrast agent image of the concentration given by the perfusion estimate, performing Monte Carlo simulations to estimate the collimation of photons or protons at multiple virtual collimator locations, and / or standard May be performed using collimator estimation to reconstruct SPECT or PET images using conventional tomography techniques. The estimated virtual perfusion image may be similar in readability and design to a SPECT or PET scan image, so a physician may be familiar with how to analyze the estimated perfusion image.

  The above-listed steps of methods 200 and 300 may be used to estimate perfusion in various tissues. Various tissues include cardiac muscle using coronary blood vessel model, brain using cerebral blood vessel model, muscle tissue using peripheral blood vessel model, liver using liver blood vessel model, kidney using renal blood vessel model, kidney using visceral blood vessel model using intestine However, the present invention is not limited thereto, and also includes spleen and pancreas using a blood vessel model for blood vessels supplying blood to a target organ in other organs.

  In one embodiment, perfusion estimates may also be used to improve blood flow simulation by using more accurate boundary conditions to perform simulation or estimation of blood flow characteristics.

  In one embodiment, the treatment plan and diagnosis virtually alter the input information (eg, vascular model, tissue model, patient specific physiological parameters, etc.) and influence the perfusion in the target tissue based on the modified input. It may be improved by predicting the

  FIG. 4 illustrates an example embodiment of a method 400 for training a machine learning algorithm to determine an estimated amount of perfusion at one or more locations in target tissue at one or more physiological conditions. Indicates The method of FIG. 4 may be implemented by server system 106 based on information, images, and data received from physician 102 and / or third party provider 104 through electronic network 101.

  In one embodiment, step 402 includes a database including patient-specific physiological parameters of one or more locations in the blood vessel and / or target tissue model and one or both of estimated or measured perfusion data at those locations. It may include assembling. “Patient-specific physiological parameters” are medical image properties received or calculated, anatomical characteristics, perfusion territory, blood supply to target tissue, blood flow characteristics, patient characteristics, disease It may refer to one or more of load characteristics and electromechanical measurements. Locations may be obtained from one or more of the available imaging or scanning modalities including, but not limited to, patient-specific vascular and / or target tissue models or PET, SPECT, MR perfusion, and / or CT perfusion. From the image being

  In one embodiment, step 404 is machine learning to map one or more patient specific physiological parameters in one or more locations of the blood vessel and / or target tissue model to perfusion data in such locations. It may include training an algorithm. Machine learning algorithms may take many forms, including, but not limited to, multi-layer perceptrons, deep learning, support vector machines, random forests, nearest neighbors, Bayesian networks, and the like.

  In one embodiment, step 406 applies the trained machine learning algorithm to a set of patient specific physiological parameters of the blood vessel and / or target tissue model obtained from the new patient, thereby: Estimating, applying perfusion data at one or more locations may be included.

  FIG. 5 is a block diagram of an exemplary method for estimating patient-specific blood flow characteristics from vascular geometry and physiological information according to an exemplary embodiment of the present disclosure. The method of FIG. 5 may be implemented by server system 106 based on information received from physician 102 and / or third party provider 104 through electronic network 100.

  In one embodiment, the method of FIG. 5 is a training method 502 for training one or more machine learning algorithms based on patient specific physiological parameters from multiple patients and measured or estimated perfusion characteristics. And, a generation method 504 may be included to use a machine learning algorithm to predict the perfusion characteristics of a particular patient.

  In one embodiment, the training method 502 comprises (a) a patient-specific geometric model, (b) one or more measured or estimated patient-specific physiological parameters, for each of a plurality of individuals. And (c) collecting values of perfusion characteristics, for example in digital form. The training method 502 may then include, for one or more points in each patient model, creating a feature vector of the patient's physiological parameters and associating the feature vector with the value of the perfusion characteristic. . The training method 502 then trains a machine learning algorithm (e.g., using the processing device of the server system 106) to perfuse at each point of the geometric model based on the feature vectors and estimated perfusion characteristics. It may be predicted. The training method 502 may then store the results of the machine learning algorithm, including the feature weights, in a storage device of the server system 106. The stored feature weights may define the degree to which patient specific physiological parameters and / or anatomical geometries predict certain perfusion characteristics.

  In another embodiment, training method 502 may be performed based on FFR estimators generated using computational fluid dynamics (CFD) for multiple patients. The training method 502 then associates the estimated FFR values with all points in the patient's geometric model, and then generates a feature vector of patient-specific physiological parameters, and a feature vector. It may include associating with FFR estimators. The training method 502 then trains a machine learning algorithm (e.g., using the processing device of the server system 106) to predict perfusion at each point in the geometric model based on the feature vectors and the estimated FFR. You may

  In one embodiment, generation method 504 may include estimating perfusion characteristics for a particular patient based on the results of performing training method 502. In one embodiment, generation method 504 may, for example, digitally capture (a) a patient-specific geometric model and (b) one or more measured or estimated patient-specific physiological parameters. You may include doing. For multiple points in the patient-specific geometric model, generation method 504 may include creating a feature vector of patient-specific physiological parameters used in the training mode. The generation method 504 may then use the saved results of the machine learning algorithm to generate an estimate of the patient's perfusion characteristics for each point in the patient-specific geometric model. Finally, the generation method 504 may include storing the results of the machine learning algorithm, including the predicted perfusion characteristics, in a storage device of the server system 106.

  Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

Claims (20)

A computer-implemented method for estimating a patient-specific value of blood perfusion through a patient's anatomy, comprising:
Receiving a patient-specific tissue model of the target tissue to be perfused by the blood vessel of a patient-specific blood vessel model and the blood vessel of a patient,
And receiving or calculating a patient-specific data of one or more of one or more raw physical parameter of the patient from the patient-specific blood vessel model or the patient-specific tissue models in physiological state of the patient ,
Wherein the patient-specific data of one or more physiological parameters, and a plurality of the one or more physiological parameters characteristic of the measured perfusion of blood through the plurality of individual tissue data of individuals and that the measured using a machine learning algorithm trained to map to the data to estimate the characteristics of the irrigation flow of blood through the patient-specific tissue model,
And outputting the pre Ki推 constant properties of irrigation flow of blood electronic display
A computer-implemented method, including : The computer-implemented method of claim 1, wherein the one or more physiological parameters comprise an estimated or measured blood flow to a plurality of regions of the blood vessel or tissue in one or more physiological conditions. The one or more physiological parameters, vessel size, blood vessel shape, vascular tortuosity, vessel length, vessel thickness, including tissue or blood of the estimation territory of irrigation flow in the vascular network, or a combination thereof, it The computer-implemented method of claim 1, comprising one or more anatomical characteristics that is not limited.   The physiological condition may include one or more of resting patient condition, congestion condition, exercise condition, postprandial condition, gravity condition, emotional condition, hypertension condition, drug taking condition, or a combination thereof. The computer implemented method according to 1. Characteristics of irrigation flow blood, the size of the flow, the direction of flow or one or more combinations thereof, computer implemented formula The method of claim 1. The one or more physiological parameters include one or more of a target tissue or blood vessel model in one or more physiological conditions, including one or more of local average intensity, texture characteristics, standard images, or a combination thereof. The computer-implemented method of claim 1, comprising a plurality of image characteristics. 10. The method of claim 1, wherein the one or more physiological parameters further comprise a plurality of secondary characteristics including one or more of patient characteristics, target tissue disease characteristics, electromechanical measurements, or a combination thereof. Computer-implemented method. Wherein estimating the characteristics of the irrigation flow blood through the patient-specific tissue models, in one or more physiological conditions comprises comparing the blood flow of the target tissue, a computer implemented as recited in claim 1 Expression method. Before Symbol patient-specific blood vessel model and the patient-specific organization model, coronary model and cardiac muscle, brain blood vessel model and brain, peripheral vascular model and muscle, liver blood vessel model and liver, renal vascular model and kidney, visceral blood vessel model and The computer-implemented method of claim 1, comprising one or more of a gut or a blood vessel model having any target organ and a blood vessel supplying the target organ. And adjusting the patient-specific data of the one or more physiological parameters based on prior Ki推 constant properties of irrigation flow blood,
And simulating the blood flow characteristics using the adjusted patient specific data to prior Ki推 constant properties and the one or more physiological parameters of irrigation flow blood
The computer-implemented method of claim 1, further comprising: Receiving one or more desired perfusion characteristics in one or more locations of the target tissue,
And that said one or estimated properties of the plurality of irrigation flow of blood in place of the target tissue is compared with the desired perfusion characteristics in the one or more locations,
Based on the comparison, the patient-specific data of the one or more physiological parameters, and changing the one or more patient-specific blood vessel model or patient-specific set of Omo del
The computer-implemented method of claim 1, further comprising: Using one or more characteristics of the estimated the irrigation flow blood, one or more virtual perfusion image is designed SPECT or PET scan in one or more physiological conditions to simulate and to estimate the,
And outputting the one or more virtual perfusion image in an electronic storage device
The computer-implemented method of claim 1, further comprising: A system for estimating patient-specific perfusion, comprising
A data storage device storing instructions for determining patient specific characteristics of blood perfusion;
Processor and
The processor is configured to perform the method by executing the instructions ;
The method is
Receiving a patient-specific vascular model of the patient's blood vessels and a patient-specific tissue model of target tissue perfused by the blood vessels ;
Receives or calculates one or more of the one or more patient-specific data of the raw physical parameter of the patient from the prior SL patient specific vessel model or the patient-specific tissue model Te physiological condition smell of the patient and the step,
Wherein the patient-specific data of one or more physiological parameters, and a plurality of the one or more physiological parameters characteristic of the measured perfusion of blood through the plurality of individual tissue data of individuals a step of measured using a machine learning algorithm trained to map to the data to estimate the characteristics of the irrigation flow of blood through the patient-specific tissue model,
And outputting the pre Ki推 constant properties of irrigation flow of blood de Isupurei
Including the system. The one or more physiological parameters may, in one or more physiological conditions,
An estimated or measured blood flow to a plurality of regions of the blood vessel or tissue;
Anatomical characteristics,
Image characteristics ,
Patient characteristics,
Target tissue disease characteristics,
Electro-mechanical measurement value, or,
To the union
The system of claim 13, comprising one or more of:   14. The physiological condition includes one or more of resting patient condition, congestion condition, exercise condition, postprandial condition, gravity condition, emotional condition, hypertension condition, drug taking condition, or a combination thereof. System described. Characteristics of irrigation flow blood, including the flow size, the one or more flow directions, or a combination thereof, The system of claim 13. Wherein estimating the characteristics of the irrigation flow blood through the patient-specific tissue model comprises comparing the blood flow before Symbol target tissue, one, or more physiological conditions odor of claim 13 system. Before Symbol patient-specific blood vessel model and the patient-specific organization model,
Coronary blood vessel model and myocardium,
Cerebrovascular model and brain,
Peripheral vascular model and muscle,
Liver blood vessel model and liver,
Renal vascular model and kidney,
Visceral vascular model and intestine , or
Vascular model having blood vessels that supply any target organ and the target organ blood
The system of claim 13, comprising one or more of: And adjusting the patient-specific data of the one or more physiological parameters based on prior Ki推 constant properties of irrigation flow blood,
And simulating the blood flow characteristics using the adjusted patient specific data to prior Ki推 constant properties and the one or more physiological parameters of irrigation flow blood
The system of claim 13, further comprising: A non-transitory computer readable medium for use on a computer system including computer executable programming instructions for estimating patient specific characteristics of blood perfusion, the computer executable programming instructions being executed by a processor Causing the processor to execute the method,
The method is
Receiving a patient-specific tissue model of the target tissue to be perfused by the blood vessel of a patient-specific blood vessel model and the blood vessel of a patient,
And receiving or calculating a patient-specific data of one or more of one or more raw physical parameter of the patient from the patient-specific blood vessel model or the patient-specific tissue models in physiological state of the patient ,
Wherein the patient-specific data of one or more physiological parameters, and a plurality of the one or more physiological parameters characteristic of the measured perfusion of blood through the plurality of individual tissue data of individuals and that the measured using a machine learning algorithm trained to map to the data to estimate the characteristics of the irrigation flow of blood through the patient-specific tissue model,
And outputting the pre Ki推 constant properties of irrigation flow of blood de Isupurei
Non-transitory computer readable media, including :

Images